Outsourcing teenage language: a participatory approach for exploring speech and text messaging

2.1 Youth language and citizen sociolinguistics

The period of adolescence is characterized by a developmental transition from childhood to adulthood, including rapid physical and cognitive changes, as well as risk-taking and novelty-seeking behaviour (Spear 2000). Teenagers’ tendencies to seek ways to differentiate themselves, and to define themselves in opposition to earlier generations, may underlie the frequent stigmatization of their language by speech communities (Labov 2001: 514). The language spoken by teenagers in English-language cultures has been investigated from various angles; it is also strikingly familiar on a global stage, thanks to the popularity of anglophone entertainment. Language change has been linked broadly to sociocultural change, and more specifically to the processes of globalization and mediatization, understood as “the pervasive spread of media contents and platforms through all types of context and practice”, that is, technically mediated communication with permanent access to media content (Androutsopoulos 2014: 7; Couldry and Hepp 2013). Mediatization has shaped the lives of the younger cohorts of today, raised in a culture characterized by omnipresent digital technology (Calvo-Porral and Pesqueira-Sanchez 2020; Kalmus and Opermann 2019). From our perspective, not only does media provide a 24/7 context for peer-oriented and meaning-making teenagers, but it also facilitates English-language contact. Estonian teenagers have frequent and intense contact with English, and participate in interactive global media communities, but the immediate social context consists in their peer groups and Estonian society (cf. Kerswill et al. 2013). This duality points to the need for much more research on the non-English sociolinguistic contexts emerging out of the globally connected media environment (Stanford 2016), to better describe and understand the factors shaping today’s multilingual teenagers’ language use and its characteristic features (e.g., vogue words, swear words, discourse markers; see Stenström et al. 2002: 64–65).

Over the last decade, sociolinguistics has joined other disciplines in pursuing citizen science, “a methodology suitable for gathering large amounts of data”, which, in the case of language studies, also challenges the notions of authority and source (Svendsen 2018: 139). Advocates of this approach in sociolinguistics highlight the unique affordance of social media behaviour in embracing speakers’ expertise and understanding the role of language in social life (Rymes 2020). The need for citizen science in variation studies comes about for (a) epistemological reasons, including efforts to counteract the neglect of mixed speech and contact languages in earlier research; (b) technological reasons deriving from developments in information and communications technology; and (c) applied reasons, accounting for the need to broaden the picture of language usage for teaching and communication (Rymes and Leone 2014). Outsourcing the collection of language data by way of citizen sociolinguistics empowers language informants by giving them the ultimate say in what is shared with researchers. It also has the virtue of simultaneously promoting the two-way street underlying best research practices, in which informants possess full knowledge of the goals and potential benefits of the research and allowing ethical access to behaviour otherwise inaccessible to out-group researchers, such as talk about intimate issues or the use of taboo language (Drange et al. 2014; Sturtz Sreetharan et al. 2019).

The citizen sociolinguistic approach is used here to reach Estonian teenagers, a demographic experiencing rapid societal and linguistic change, whose language is strikingly underrepresented in corpus data. Millennial and post-millennial cohorts in Estonia, as elsewhere, are both experiencers and promoters of an influx of English and the rise of digital technology as primary communication channels. The easily accessible global market and discourse space has brought with it the relativization of values and relaxation of restrictions, as well as a countervailing rise in prescriptivist language attitudes.

2.2 Research into youth language in Estonia

Scholarly interest in youth language in Estonia can be traced to the mid-1920s, when Andrus Saareste, extraordinary professor of linguistics at the University of Tartu, called for investigation of social variants. While Saareste (1927: 167) considered youth especially prone to language innovation, he also warned that not all innovations were welcome and considered the use of foreign loans to be affective rather than “rational”. His views corresponded to the dominant stance of the era and set the tone in research for the coming decades.

Generally, however, youth language remained a marginal research subject until the 1970s and 1980s, when Labovian urban dialectology reached Estonian linguists.^[2] Even then, youth language research remained situated within slang studies or focused on urban youth for decades (Loog 1992; Peegel 1976; Tender 1984). Data collection methods included word lists, known from (descriptive) dialectology, written questionnaires, and interviews where informants were asked to respond to stimuli (e.g., Keevallik 1998). Only one study aimed at collecting the unmonitored speech of urban youth, involving secondary students in the data collection (see Loog 1988a, 1988b).

While the corpus of spoken Estonian was launched in 1997 (Hennoste 2003), that corpus does not include any samples dedicated to teenagers. This dearth of teenage language samples was true in other languages as well: the last three decades have seen the compilation of youth language corpora in English (e.g., COLT, see Stenström et al. 2002; Toronto Teen Corpus, Toronto Instant Messaging Corpus, Tagliamonte 2016); Spanish (COLA, Stenström 2006); French (Secova 2017); and Portuguese (Mata et al. 2014). In the Estonian context, the project described here is thus far the only one devoted to the compilation of a corpus on this scale. A bilingual English–Estonian corpus mainly comprising the language of bloggers and vloggers, mostly in their twenties, has also recently been assembled (Bahtina et al. 2021). However, no representative corpus of teen language has been available for systematic study.

3.1 Recruitment and training

We recruited participants through schools, supplemented by personal contacts to redress imbalances and target groups who proved more difficult to access. Targeted students were in fourth, sixth, eighth, and tenth grade, which in Estonia spans the ages of 10–17. The youngest, pre-teen group was a smaller sample, included for apparent-time comparison, as we expected them to show fewer characteristic features of teenage language use, and to talk about different topics from the older groups (cf. Stenström et al. 2002: 28–29). To compare language use across regions differing in the presence of languages other than Estonian and its dialects, four locations were selected. First, we chose the capital city, Tallinn (approximately 438,000 inhabitants), where residents include Estonian and Russian speakers in similar proportions, and more tourists and migrants than elsewhere in Estonia. Second, we selected Tartu, Estonia’s second largest city (about 98,000 inhabitants), where standard Estonian is spoken as the dominant native language. Finally, we included one smaller town where the insular dialect is spoken, and another in the South Estonian dialect area.^[3] In each location, we targeted a gender-balanced sample, with two fourth graders and four participants from each of the older age groups (sixth, eighth, and tenth grades), for a total of 14 lead informants from each location, or 56 informants altogether, who additionally brought their friends into the project.

We first contacted teachers and heads of target schools via email. After positive replies from schools, we scheduled school visits to introduce the project and recruit interested students as lead participants, or ‘language ambassadors’ (keelesaadikud in Estonian). However, the outbreak of COVID-19 and subsequent lockdown led us to re-envision the recruitment and training process, using an entirely online format. We created a 5-min video [associated video-1-koreinik.mp4] describing the project, discussing the value of young people’s language for research and society, and inviting interested students to send us their contact details, personal interests, and reasons for wanting to join the project. We asked teachers to share the video among their students and posted it on the project website. Some teachers declined, reporting that their students were already challenged by the transition to remote learning (see Section 3.2), but others felt that participation in this project was an excellent opportunity for a rare, permissible interactive experience, as it involved pairing students with only one or two friends who were part of the student’s social contact “bubble”. Those teachers transmitted the project invitation, and in the first, springtime round we received applications from various age groups and regions, totalling 18 applicants.

Applicants were invited to participate in two 45-minute online training sessions via Zoom, run on different days by members of the research group. The first training session was used to inform trainees about the project, to encourage reflection and mediate discussion about trainees’ own language use, and to discuss ethical issues in collecting language data, including trust, privacy, and data accessibility. We explained the need for explicit, written consent from participants and their parents. The second training session focused on technical instruction in recording spoken language using audio recorders, accessing and submitting chat conversations and the required metadata, and other necessary steps for completing the project. Trainees were also given a homework assignment between the two sessions involving asking for consent and recording a couple minutes of conversation with a family member, listening to the recording, and responding to discussion questions. Their experiences and observations sparked an often spirited discussion at the second session.

Once they had completed training, the language ambassadors were given two primary tasks: (a) recording and submitting spontaneous spoken conversations and (b) submitting conversations held over text messaging applications, the most popular of which were Facebook Messenger and Discord. As the ambassadors and their friends were minors, ethics approval was secured from the Research Ethics Committee of the University of Tartu before contacting schools, and informed consent was obtained from all participants (including their friends and interlocutors in texted conversations) and their parents or guardians.

For the spoken language samples, participants were asked to recruit at least one friend, or co-informant, who would be willing to engage in an hour-long spoken conversation, and to organize and record two such conversations. Zoom H2N audio recorders were provided by the research group via a parcel delivery service, as direct contact through schools was hampered during lockdown. After the recordings were completed, the recorders and their memory cards were collected directly from the informants or teachers.

For the texted chats, informants were asked to look for digital texted conversations which they were willing to share with the research team, for a minimum of 1,000 words of conversation. We took a retrospective approach, advising participants to collect preferably pre-existing conversations in order to reduce effects of the project in the chats. We aimed to collect 112 h of spoken language and a minimum of 56,000 words of text message data. Contextual metadata on the recordings (e.g., participants, recording situation, place, and time) was collected through an online form. Background data on each participant in the conversations was collected via an online survey asking for school, place of residence, language skills and use, and social media use. Forms for collecting metadata and instructions on how to use the project recorders and how to download and submit text messages were available on the project website during data collection. Instructions in Estonian are now accessible on the page for students and teachers (https://sisu.ut.ee/teke/opetajale_ja_opilasele), under the heading TeKE projektis kasutatud ankeedid ja juhendid ‘Documents and instructions used in the TeKE project’, along with information on outreach classes dedicated to youth language and linguistics.

Participants were compensated for their time. The ambassadors were given a monetary stipend upon completion of all the required tasks, and each participant received a gift card per one hour-long spoken conversation (ambassadors received gift cards in addition to the stipend).

3.2 Remote data collection during COVID-19 restrictions

The outbreak of the COVID-19 pandemic considerably affected the data collection process. As Estonia implemented a full lockdown in mid-March, just as our data collection was meant to begin, we were unable to make any face-to-face contact with participants. The lack of direct contact made recruitment more difficult, and various teachers reported difficulty in contacting students remotely. Some parents reported concerns about their children’s sense of being overwhelmed, and this topic arises in the recordings as well. In addition, it proved to be more difficult to reach students online than at a face-to-face project introduction in the classroom, as initially planned. For the students, contacting the research group by email is likely to have been more of an obstacle than a question-and-answer session at the school. Online, written instructions via email and the internet were more difficult for students to follow than direct instruction in classrooms, especially for younger participants.

All participants were encouraged to contact the project members with any questions or concerns, and the project coordinator monitored their progress closely and offered assistance where possible. In the case of younger participants, some parents took a more active role in helping their children and contacting the project representatives. The natural filter this provided likely impacted on the sample of participants, those joining the project having already demonstrated an amount of drive and persistence, in addition to the dedication and responsibility required to attend and actively participate in both online training sessions.

It was clear that the lockdown affected the participants, especially due to a lack of direct, in-person contact with their friends and teachers. Some participants reported difficulties in arranging meetings with their friends for recording conversations; in a small minority of instances, we allowed the option of recording conversations held online, over a video call. As could be expected, the pandemic and concomitant lockdown had profound effects on the topics discussed in conversations. Due to these challenges, both recruitment of participants and the data collection process were delayed and slowed down.

Yet the acute sense of living through an extraordinary experience during the pandemic may also have fuelled interest in the project, as an outlet for productive energy, a diversion from the stress of remote schooling, and a channel for communicating socio-emotional difficulties or demonstrating the frustrations and sense of isolation of the pandemic era.

3.3 Engaged participants, collected data

After the first trial wave of participants in April and September of 2020, the project undertook a second recruitment wave at the end of 2020, when schools were partially engaged in contact teaching and social life had some semblance of normalcy, although both individuals and entire schools were in isolation for parts of the 2020–2021 academic year as well. From our first wave, we had amassed experience in how to run the recruitment, training, and monitoring over the internet, and the second wave proved successful in recruiting a pool of participants similar to the original aims. A total of 47 ambassadors completed the tasks (with 56 originally targeted); see Figure 1.

Figure 1:

Participants (ambassadors and total participants) in the corpus, by age and gender.

Gender balance was not achieved: the set of ambassadors included over twice as many girls as boys (32 female, 15 male). Each ambassador invited one or more friends to participate in the spoken and text conversations; here, the gender imbalance became even more pronounced: a total of 84 friends joined the participant sample, with 62 females and 22 males. Altogether, the corpus includes language use from 131 participants (72 percent female), ranging in age from 9 to 18 (mean = 13.68, SD = 2.40).

Informants provided a total of 97 h and 13 min of spoken conversation and 59,688 words of texted conversations, surpassing the minimal expectation of 56,000 words. No upper limit on texted conversations was set, and many informants contributed more than 1,000 words, with the notably large contributions of two sixth graders who sent 9,346 and 37,288 words.^[4] Setting aside these two outliers, the agents sent a mean of 1,296 words each (SD = 991). Although our minimum of 1,000 words was specifically intended to make it easier for less text-friendly participants to fulfil the task, some younger participants were not as used to chatting with their friends and seemed to devise chats especially for project purposes, as in the two lines in (1a) and (1b), written by a fourth-grade girl to her best friend.

A team of mostly student transcribers are, at the time of writing, in the process of transcribing the spoken conversations using ELAN software,^[5] according to transcription guidelines worked out by the project team. A user-friendly application is nearly complete for help with easy, parallel searches of both corpora, but the texted chats require more manual tagging to make them accessible and searchable in a similar way. As of 1 May 2023, about 78 h of spoken conversation had been transcribed, amounting to 646,351 words. The transcription process includes coding for language used, pseudonymization, and omission of identifying information. To provide access to the data, transcriptions and metadata will be deposited in the University of Tartu’s DataDOI repository. Stable versions of the corpus will be assigned digital object identifiers (DOIs) in order to facilitate searching and to encourage the reuse of the data. Because the data derives from the spontaneous language of minors, and hence requires protected access, researchers interested in the corpus will be able to apply for access to the corpus.